Assessment-first
The goal is modernization planning, not continuous metadata management.
Methodology
Dependency evidence, not guaranteed lineage.
The assessment is designed to be useful without pretending to be omniscient. Confidence, coverage, and caveats are part of the output.
Evidence types
Direct metadata, parsed definitions, command text, query parse, artifact parse, and heuristic inference.
Confidence levels
High for direct or literal evidence, medium for strong parse evidence, low for inferred or ambiguous links.
Coverage status
Full, partial, minimal, or unavailable coverage should be visible at source and asset level.
Complexity scoring
Use a 1-5 external scale with drivers, caveats, and missing-information context.
Human review
Automation accelerates discovery; architecture interpretation still matters for roadmap decisions.
How the methodology works
Evidence is organized before recommendations are discussed.
How it works explains the operational flow. Methodology explains how the output should be interpreted: what is known, what is inferred, what is missing, and where consultant judgment is required.
1. Separate facts from interpretation
Inventory, scan status, raw artifacts, and direct metadata are kept distinct from inferred relationships and recommendation signals.
2. Attach confidence to relationships
A dependency is more useful when a reviewer can see whether it came from direct metadata, parsed source, or a weaker inference.
3. Score with caveats
Complexity and risk signals stay simple enough for planning, but include drivers and missing-information pressure.
4. Turn evidence into options
The readout compares retain-and-investigate, phased modernization, priority planning, and open questions rather than declaring an automatic target platform.
How caveats are handled
The assessment does not claim perfect lineage. It produces dependency evidence from available metadata, parsed artifacts, execution signals, and heuristics. Permissions, unsupported versions, dynamic SQL, expressions, retained history, and masked values can reduce coverage and confidence. Those limitations are shown in the output rather than hidden.
Next step
Use the methodology to frame the readout.
The methodology is there to keep recommendations honest: evidence first, confidence visible, caveats explicit, and roadmap decisions reviewed by people.